It'd be good to add a comment above this saying something like:

// Support comma-separated selectors by exploding them into an array.

I guess it's obvious in the context of this PR given it mentions the various heading block tags, but it might not be so obvious if you're just looking at this file on its own wondering why it works this way.

Similarly, it might be worth a comment saying something like:

// Creating a new tag processor on every loop resets the search to the start.

Not really sure these comments are necessary, seeing how the same thing is listed below in an easy to read format. 😄

Unfortunately, button blocks don't seem to be working correctly. When they have an override they're missing the <div class="wp-block-button"> wrapper.

It's because the code below does this:

$tag_name     = $tags->get_tag();
$markup       = "<$tag_name>$custom_value</$tag_name>";
$updated_tags = new WP_HTML_Tag_Processor( $markup );
$updated_tags->next_tag();
return $updated_tags->get_updated_html();

So it considers the updated html of the block to just be the tag that it's updating and disregards any markup before or after that.

Instead it needs to update the $original_tags and get_updated_html from that, but I don't think it's supported yet 🤔

It is mentioned in the block bindings tracking issue - #54536.

Support to mutate HTML to change the inner content of an element. Until we have a set_inner_markup function, we can manually mutate the HTML for our experiments.

I guess we need to drop button support until then.

Image also doesn't work, as the code only updates tag content, not attribute values, so you end up with <image>$url</image> and it would also strip the image block <figure> wrapper, so I think that should be removed too.

@SantosGuillamot opened #57249 that updates the PHP implementation for block bindings that should fix the handling for the Heading and Button blocks. It was extracted from the full prototype for Custom Fields: #56867.

we could invert this loop and avoid the confusion @talldan wants to clarify, while improving performance.

• iterate through all the tags until
• we find a tag matching one of the selectors

this assumes that there are likely more tags than selectors, many more, and so the cost of checking each selector at each tag will be cheaper than checking every tag at each selector. additionally, we should make it pretty explicit here in a comment that the only selectors we currently support are tag names.

$selectors = array();
foreach ( explode( ',', $block_type->attributes[ $attribute_name ]['selector'] ) as $tag_name ) {
	$selectors[] = trim( strtoupper( $tag_name ) );
}

$tags = new WP_HTML_Tag_Processor( $block_content );
while ( $tags->next_tag() ) {
	$tag_name = $tags->get_tag();
	
	if ( ! in_array( $tag_name, $selectors, true ) ) {
		 continue;
	}
	
	$markup       = "<$tag_name>$custom_value</$tag_name>";
	$updated_tags = new WP_HTML_Tag_Processor( $markup );
	$updated_tags->next_tag();

	// Get all the attributes from the original block and add them to the new markup.
	$names = $tags->get_attribute_names_with_prefix( '' );
	foreach ( $names as $name ) {
		 $updated_tags->set_attribute( $name, $tags->get_attribute( $name ) );
	}

	return $updated_tags->get_updated_html();
}

return $block_content;